About me

I am a Ph.D. student at the Department of Automation, Shanghai Jiao Tong University (SJTU), working with Prof. Cailian Chen, Prof. Jianping He, and Prof. Xinping Guan. I received the B.Eng. and M.S. Eng. degrees in Automation from North China Electric Power University in 2016 and 2019, respectively.

I am focusing on Machine Learning for Data-driven Modelling, Safety-critical, and Optimal Control. I dedicate to applying these techniques to Digital-Twin Systems for industry and industrial robotics.

What i'm doing

Working Partners



  1. North China Electric Power University

    2012 — 2016

    Received a B.Eng degree in Measurement & Control Technology and Instrumentation.

  2. North China Electric Power University

    2016 — 2019

    Received an M.S.Eng degree in Pattern Recognition & Intelligent Systems.

  3. Shanghai Jiao Tong University

    2019 — Present

    Pursuing a Ph.D. degree in Control Science and Engineering.


  1. Safety-critical Control in Motion Planning (Oxford)

    2022.09 — Present

    Designing a safety-critical control method is fundamental for robotics. By exploring the control invariance, we designed a CBF-based QP control method to guarantee the dynamics' safety intrinsically. We mainly focus on the following: 1) Controller construction method based on environment and dynamics; 2) Feasibility of the controller in a cluttered environment. 3) High computing time of the controller; Together, the online optimal control for safety-critical robotics is efficient.

  2. Intelligent Control System with Virtual-Physical Interaction for Robots (Tencent)

    2022.06 — 2022.09

    I worked with Robotics X to develop a digital-twin system for robotics, including wheeled and quadruped walking robots. The system aims to verify strategies and present a virtual-reality environment. I focused on 1) developing a cross-platform communication framework with low latency; 2) constructing robot models and control in Unreal Engine.

  3. Robotics Platform with Digital Twin Technique

    2021.03 — 2022.06

    Designed and established a novel multi-robot testbed that exploits the ideas of virtual-physical modeling in a digital-twin system. Unlike most existing testbeds, the developed Robopheus constructs a bridge that connects traditional physical hardware and virtual simulation testbeds, providing both sides with scalable, interactive, and high-fidelity simulations-tests. Another salient feature of the platform is that it enables a new form to dynamically learn the actual models from the physical environment and is compatible with heterogeneous robot chassis and controllers. In turn, the virtual world's learned models are further leveraged to approximate the robot dynamics online on the physical side. The self-designed Omni-directional robot can achieve a stable and accurate movement of the robot by coordinating the control of various actuators (velocity-tracking mode). The self-designed low-cost 2-D motion caption system can achieve 100Hz tracking frequency, 10ms delay, and 0.5mm tracking error.

  4. Low Cost Autopilot Field Vehicle

    2020.12 — 2022.02

    I designed an autopilot field vehicle with low-cost hardware which can autopilot to the desired position with obstacle avoidance. There are two visions of the system. 1) 8-bit MCU with GPS, IMU, and ultrasound detectors. The positional accuracy is improved by fusing information from GPS and IMU. The fusion algorithm is modified to adapt to the 8-bit MCU. The positional error is 2m, and the frequency is 20Hz. 2) 32-bit ARM Cortex-M3 with Dual-GPS, IMU, magnetic sensor, encoders, and ultrasound detectors. The positional error is 0.8m, and the frequency is 20Hz.

  5. Sintering Digital Twin System (Liugang)

    2019.12 — 2020.12

    I designed the architecture of the sintered digital twin platform. The platform aims to provide process variable prediction, strategy optimization, and verification. Microservices are used as the underlying driver of the platform. Docker technology is used for cross-platform packaging models, and harbor storage and K8s management are used to support multi-model twin platforms. I designed about 80 lightweight data-driven sintering quality prediction models with different functions. More than 200 microservices have been developed to cooperate with the prediction models to predict different variables and ensure the stable operation of the twin platform. I deployed the model to the twin platform and completed the development of the front-end display page.

  6. NOx prediction in coal-fired power plant(NCEPU)

    2017.01 — 2018.12

    The proposed data-driven NOx prediction method includes on-site data SCADA acquisition and processing, information entropy, and HITS algorithm for practical information data mining, designing LSSVM update mechanism, and prediction of NOx content. Control logic building and on-site DCS configuration and debugging, using data analysis results, design control logic, setting up system parameters, and completing the control system upgrade on the spot.

My skills

  • Optimal Control
  • Motion Plan
  • Stastical Learning
  • C++
  • Python


Safe Control For Manipulators

Optimal Motion Plan

Camera Sensing Model


Contact Form